Overview of Tourism Trend and Corresponding Factors

Author: Xiangzhi Chen, Zenan Wang, Zihao Huang, Jiachen Gao

From: Georgetown University

Introduction

Tourism has long been a powerful force in shaping the world—connecting people across continents, fostering cultural understanding, and driving economic growth. In recent decades, global tourism has evolved into a major pillar of the world economy, influencing everything from employment rates to infrastructure development. However, tourism is more than just a leisure activity; it is deeply intertwined with socioeconomic factors such as GDP growth and workforce dynamics. Understanding these complex relationships is crucial.

This project, developed at Georgetown University, aims to guide audiences through an analytical journey of global tourism trends. We hope to explore how tourism patterns have changed over time and across regions. By investigating the correlations between tourism and key socioeconomic indicators—such as national GDP and employment—we seek to uncover the hidden drivers behind the movement of people across the globe. Our visualizations aim to not only inform but also inspire a deeper appreciation for the intricate forces shaping the world of travel today.

The Data We Use

To conduct our analysis, we primarily draw from two rich data sources. The first is a dataset from the United Nations World Tourism Organization (UNWTO), which provides comprehensive information on international tourist arrivals across different countries and regions over time. This dataset serves as the foundation for identifying overarching tourism trends and regional disparities.

dataset 1

In addition, we supplemented our analysis with socioeconomic data sourced from the World Bank database. By carefully selecting indicators such as GDP per capita and employment rates, we built a multidimensional view of the factors influencing tourism. This blended approach allows us to not only track tourism flows but also investigate the economic conditions that might be driving tourist movements globally.

dataset 2

Global Tourism Trend

Given the vast number of countries and the richness of global tourism patterns, it was important for us to first step back and look at the world as a whole. We created an interactive global map with a dropdown button for inbound tourism and domestic tourism, covering data from 2010 to 2022. This plot aims to guide our country’s selection for deeper analysis. Each bubble on the map represents a country, with the size of the bubble proportional to the total number of arrivals. Countries with stronger tourism activity naturally stand out with larger, more vibrant bubbles.

From these visualizations, several countries clearly emerged as leaders:

• United States and China: Strong both in domestic and inbound tourism.

• United Kingdom: Massive domestic tourism and notable international activity.

• France and Spain: Consistently high inbound tourist numbers, highlighting their global appeal.

• India: Surprisingly large domestic tourism.

Since it would be impractical to analyze every nation individually, we strategically focused on these countries, which represent a variety of tourism dynamics: mature tourism markets, emerging tourism powers, and countries with interesting contrasts between domestic and international travel patterns.

Code

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import matplotlib.style as style
import plotly.express as px

import pycountry
import plotly.graph_objects as go

# Load data
arrival = pd.read_csv("../data/Processed_data/arrival.csv")
domestic = pd.read_csv("../data/Processed_data/domestic.csv")

# Filter data by year (2010 - 2022)
arrival_2010_2022 = arrival[(arrival['Years'] >= 2010) & (arrival['Years'] <= 2022)]
domestic_2010_2022 = domestic[(domestic['Years'] >= 2010) & (domestic['Years'] <= 2022)]

# Summarize total arrivals by country
total_arrivals_by_country = arrival_2010_2022.groupby('Country', as_index=False)['Total arrivals (Thousands)'].sum().rename(columns={'Total arrivals (Thousands)': 'total_arrivals'})

total_domestic_by_country = domestic_2010_2022.groupby('Country', as_index=False)['Total trips (Thousands)'].sum().rename(columns={'Total trips (Thousands)': 'total_arrivals'})

# Ensure numeric and handle missing values
total_arrivals_by_country['total_arrivals'] = pd.to_numeric(total_arrivals_by_country['total_arrivals'], errors='coerce').fillna(0)
total_domestic_by_country['total_arrivals'] = pd.to_numeric(total_domestic_by_country['total_arrivals'], errors='coerce').fillna(0)

# Filter top countries
top_countries_arrival = total_arrivals_by_country.nlargest(30, 'total_arrivals')
total_arrivals_by_country['country_group'] = total_arrivals_by_country['Country'].apply(
    lambda x: x if x in top_countries_arrival['Country'].values else 'Other'
)

top_countries_domestic = total_domestic_by_country.nlargest(30, 'total_arrivals')
total_domestic_by_country['country_group'] = total_domestic_by_country['Country'].apply(
    lambda x: x if x in top_countries_domestic['Country'].values else 'Other'
)

# Add Type col
total_arrivals_by_country['Type'] = 'Inbound'
total_domestic_by_country['Type'] = 'Domestic'

# Combine both datasets
combined_df = pd.concat([
    total_arrivals_by_country[['country_group', 'total_arrivals', 'Type']],
    total_domestic_by_country[['country_group', 'total_arrivals', 'Type']]
])

# Group by Type and country_group to sum values
combined_df = combined_df.groupby(['Type', 'country_group'], as_index=False)['total_arrivals'].sum()

# Rename specific country names
combined_df['country_group'] = combined_df['country_group'].replace({
    'UNITED STATES OF AMERICA': 'US'
})

# Get ISO alpha-3 codes
def get_iso3(country_name):
    try:
        return pycountry.countries.lookup(country_name).alpha_3
    except:
        return None

# Add ISO code
combined_df['iso_alpha'] = combined_df['country_group'].apply(get_iso3)
plot_df = combined_df.dropna(subset=['iso_alpha'])

# Ensure numeric
plot_df['total_arrivals'] = pd.to_numeric(plot_df['total_arrivals'], errors='coerce').fillna(0)

fig = go.Figure()

types = plot_df['Type'].unique()

# Normalize size and enforce minimum bubble size
min_size = 10
max_size = 50

for i, t in enumerate(types):
    df_t = plot_df[plot_df['Type'] == t]
    
    max_val = df_t['total_arrivals'].max()
    normalized_size = df_t['total_arrivals'] / max_val * max_size
    size_with_min = normalized_size.clip(lower=min_size)  # Ensure minimum size since unequally distributed lol

    fig.add_trace(go.Scattergeo(
        locations=df_t['iso_alpha'],
        locationmode='ISO-3',
        text=df_t['country_group'],
        hovertext=df_t['country_group'] + "<br>" + df_t['total_arrivals'].round(1).astype(str) + "K",
        marker=dict(size=size_with_min, color=df_t['total_arrivals'],
            colorscale='Viridis', colorbar_title='Arrivals'
        ),
        name=t,
        visible=(i == 0)
    ))

# Dropdown selection
fig.update_layout(
    updatemenus=[dict(
        buttons=[dict(label=t, method='update', 
        args=[{'visible': [t == ty for ty in types]}, 
        {'title': f'Total Arrivals by Countries from 2010 to 2022 ({t})'}]) for t in types],
        direction='down', x=0.5, xanchor='center', y=1.1, yanchor='top')],
    geo=dict(projection_type="natural earth"),
    title='Total Arrivals by Countries from 2010 to 2022 (Inbound)',
    margin=dict(t=100, l=0, r=0, b=0)
)

fig.show()

Figure 1: Global Map

While global arrival numbers give us a big-picture view, understanding where tourists are coming from offers even deeper insights. To explore this, we created an interactive Sankey diagram that visualizes the flow of tourists from broader regions to specific countries. Each flow in the Sankey plot represents the volume of tourists traveling from a given region to the major countries with large arrivals. By adjusting the year, users can observe how these flows have evolved from 2010 to 2022.

Several global patterns clearly emerge:

• Tourists often stay within their own continent:

Travelers in Europe frequently visit other European countries such as France, Spain, and the UK. And in East Asia and the Pacific, regional travel is strong, with many tourists choosing China, Japan, Thailand, and Malaysia as destinations.

• Regional Leaders:

The United States stands out not only as a domestic tourism giant but also as a major destination for travelers from East Asia, the Pacific, Europe, and the Americas.

Through this visualization, we observe that geographic proximity, economic ties, and cultural familiarity heavily influence international travel choices. The Sankey plot not only helps highlight major tourism hubs but also uncovers how interconnected different regions are in the global tourism network.

Code

regions = pd.read_csv("../data/Processed_data/regions.csv")
regions.rename(columns=lambda x: x.replace(' (Thousands)', ''), inplace=True)

# Set of countries and regions
countries_of_interest = [
    'CHINA', 'UNITED STATES OF AMERICA', 'FRANCE', 'UNITED KINGDOM',
    'SPAIN', 'INDIA', 'MEXICO', 'ITALY', 'POLAND', 'JAPAN', 'THAILAND', 'MALAYSIA', 'CANADA', 'SOUTH AFRICA'
]

region_columns = [
    'Africa', 'Americas', 'East Asia and the Pacific',
    'Europe', 'Middle East', 'Other not classified', 'South Asia'
]

# Filter for relevant countries
regions = regions[regions['Country'].isin(countries_of_interest)]
years = sorted(regions['Years'].unique())

# Prepare static node list
nodes = region_columns + countries_of_interest
node_map = {node: i for i, node in enumerate(nodes)}

# Build one Sankey trace per year
data_traces = []
dropdown_buttons = []

for i, year in enumerate(years):
    df_year = regions[regions['Years'] == year]
    df_agg = df_year.groupby('Country')[region_columns].sum().reset_index()
    df_long = df_agg.melt(id_vars='Country', var_name='Region', value_name='Value')
    df_long = df_long[df_long['Value'].notna() & (df_long['Value'] > 0)]
    df_long['Value'] *= 1000  # Convert from 'Thousands' to actual numbers

    trace = go.Sankey(
        visible=(i == 0),
        node=dict(
            pad=15,
            thickness=20,
            line=dict(color='black', width=0.5),
            label=nodes
        ),
        link=dict(
            source=df_long['Region'].map(node_map),
            target=df_long['Country'].map(node_map),
            value=df_long['Value'],
            hovertemplate='From %{source.label} to %{target.label}<br>Value: %{value:,}',
            color='rgba(169, 169, 169, 0.5)'
        )
    )
    data_traces.append(trace)

    dropdown_buttons.append(dict(
        label=str(year),
        method='update',
        args=[
            {'visible': [j == i for j in range(len(years))]},
            {'title': f'Migration Flow from Regions to Countries in {year}'}
        ]
    ))

# Create the figure
fig = go.Figure(data=data_traces)
fig.update_layout(
    title=f'Migration Flow from Regions to Countries in {years[0]}',
    width=740,
    height=500,
    plot_bgcolor='rgba(0,0,0,0)',
    paper_bgcolor='rgba(0,0,0,0)',
    updatemenus=[dict(
        active=0,
        buttons=dropdown_buttons,
        x=1.1,
        y=1,
        xanchor='right',
        yanchor='top'
    )]
)

fig.show()

Figure 2: Sankey Diagram

To closely examine how inbound and domestic tourism evolved over time, we created stacked plots with 4-year rolling averages for six major countries. This approach highlights not only the overall growth trends but also the disruptions caused by global events like the COVID-19 pandemic.

When comparing across all six countries, clear patterns emerge regarding tourism structure and resilience:

China, India, the United States, the United Kingdom, France, and Spain all show that domestic tourism is the dominant force in their tourism industries. In each case, domestic trips are consistently larger than inbound arrivals across the entire period from 2010 to 2022.

China, India, and the United States display extremely large domestic sectors, where inbound tourism contributes only a very small share. The United Kingdom experienced a particularly strong domestic tourism surge post-2014, culminating in record highs by 2022.

France and Spain, while traditionally seen as major inbound destinations, also have stronger domestic markets than inbound. Domestic trips in both countries consistently outnumber inbound arrivals, although inbound tourism still plays an important complementary role.

Pandemic impacts were seen across all countries, but those with stronger domestic bases — like the United Kingdom, India, and China — exhibited faster and stronger recovery trajectories.

Across all countries, a robust domestic tourism sector proved critical for resilience during global disruptions, highlighting its importance not just for economic recovery but for the long-term sustainability of the tourism industry.

Code

import pandas as pd
import matplotlib.pyplot as plt

# Load the data
arrival = pd.read_csv("../data/Processed_data/arrival.csv")
domestic = pd.read_csv("../data/Processed_data/domestic.csv")

# Filter for the years 2010 to 2022
arrival_filtered = arrival[(arrival['Years'] >= 2010) & (arrival['Years'] <= 2022)]
domestic_filtered = domestic[(domestic['Years'] >= 2010) & (domestic['Years'] <= 2022)]

# Filter data for interested countries
countries_of_interest = ['CHINA', 'UNITED STATES OF AMERICA', 'FRANCE', 'UNITED KINGDOM', 'SPAIN', 'INDIA']
arrival_filtered = arrival_filtered[arrival_filtered['Country'].isin(countries_of_interest)]
domestic_filtered = domestic_filtered[domestic_filtered['Country'].isin(countries_of_interest)]

# Calculate 4-year rolling average for arrivals
arrival_filtered['Total_Arrival_Rolling_Avg'] = arrival_filtered.groupby('Country')['Total arrivals (Thousands)']\
    .rolling(window=4, min_periods=1).mean().reset_index(level=0, drop=True)

# Calculate 4-year rolling average for domestic arrivals
domestic_filtered['Total_Trips_Rolling_Avg'] = domestic_filtered.groupby('Country')['Total trips (Thousands)']\
    .rolling(window=4, min_periods=1).mean().reset_index(level=0, drop=True)

# Merge both arrival and domestic data
merged_data = pd.merge(
    arrival_filtered[['Country', 'Years', 'Total arrivals (Thousands)', 'Total_Arrival_Rolling_Avg']],
    domestic_filtered[['Country', 'Years', 'Total trips (Thousands)', 'Total_Trips_Rolling_Avg']],
    on=['Country', 'Years'], how='inner'
)

# Create a 2x3 grid of subplots
fig, axes = plt.subplots(2, 3, figsize=(10, 5))  # Smaller overall size
axes = axes.flatten()

for idx, country in enumerate(countries_of_interest):
    ax = axes[idx]
    country_data = merged_data[merged_data['Country'] == country]

    if country_data.empty:
        print(f"No data available for {country}. Skipping plot.")
        continue

    # Stacked bar plots
    ax.bar(country_data['Years'], country_data['Total arrivals (Thousands)'], label='Inbound Arrivals', color='#FF91A4', width=0.5)
    ax.bar(country_data['Years'], country_data['Total trips (Thousands)'], bottom=country_data['Total arrivals (Thousands)'], label='Domestic Trips', color='lightblue', width=0.5)

    # Lines of rolling averages
    ax.plot(country_data['Years'], country_data['Total_Arrival_Rolling_Avg'], label='Inbound Rolling Avg', linestyle='--', marker='x', color='#C40234', linewidth=2.5)
    ax.plot(country_data['Years'], country_data['Total_Trips_Rolling_Avg'], label='Domestic Rolling Avg', linestyle='-', marker='o', color='blue', linewidth=2.5)

    ax.set_title(f'{country}', fontsize=13)
    ax.set_xlabel('Year', fontsize=11)
    ax.set_ylabel('Tourists (Thousands)', fontsize=11)
    ax.set_xticks(country_data['Years'])
    ax.tick_params(axis='x', rotation=45)

# Handle the legends: Move to bottom center
handles, labels = axes[0].get_legend_handles_labels()
fig.legend(handles, labels, loc='lower center', ncol=4, fontsize=10, bbox_to_anchor=(0.5, -0.05))

# Adjust layout
plt.tight_layout(rect=[0, 0.05, 1, 0.95])  # Leave space for the title and legend
fig.suptitle('Inbound and Domestic Tourist Trends (2010–2022)', fontsize=18, y=0.98)
plt.show()

What Guides Tourism

GDP on Tourism: Not Really

To further explore the relationship between countries’ economic strength and their tourism performance, we animated a bubble plot showing Tourism Expenditure vs GDP from 2000 to 2022.

Each bubble represents a country, with:

• X-axis: Total tourism expenditure (in USD)

• Y-axis: GDP (in USD)

• Bubble size: Number of arrivals

As the animation plays across the years, a few key insights become clear:

• Economic factors like GDP does not appear to be a direct driver of tourism:

Countries like France, Spain, and the United Kingdom consistently have high tourism expenditures despite their moderate GDP compared to giants like the United States and China.

• The United States stands out:

It is an exception where both GDP and tourism expenditure are extremely high, suggesting that for some destinations, economic size does help boost tourism spending.

• China’s Position:

Although China’s GDP grows rapidly during this period, its tourism expenditure remains relatively more modest.

• Spain and France punch above their weight:

Despite having GDPs much smaller than China or the U.S., these countries attract huge tourism spending, reinforcing the idea that cultural and historical appeal outweighs pure economic power.

Overall, “Tourists Chase Experiences, Not Really Economies”. GDP size doesn’t guarantee tourism success.

Code

# Load the data
arrival = pd.read_csv("../data/Processed_data/arrival.csv")
expenditure = pd.read_csv("../data/Processed_data/expenditure.csv")

# Select time range
arrival_2010_2022 = arrival[(arrival['Years'] >= 2000) & (arrival['Years'] <= 2022)]
expenditure_2010_2022 = expenditure[(expenditure['Years'] >= 2000) & (expenditure['Years'] <= 2022)]

# Interested countries to select
countries = ['CHINA', 'UNITED STATES OF AMERICA', 'FRANCE', 'UNITED KINGDOM', 'SPAIN', 'INDIA']
# Filter by interested countries
arrival_selected = arrival_2010_2022[arrival_2010_2022['Country'].isin(countries)]
expenditure_selected = expenditure_2010_2022[expenditure_2010_2022['Country'].isin(countries)]

# Rename columns
arrival_selected = arrival_selected.rename(columns={'Total arrivals (Thousands)': 'Total_arrival'})
expenditure_selected = expenditure_selected.rename(columns={
    'Tourism expenditure in the country (US$ Millions)': 'Total_expend',
    'Passenger transport (US$ Millions)': 'Passenger_expend',
    'Travel (US$ Millions)': 'Travel_expend'
})

# Ensure numeric and handle missing values (for both datasets)
arrival_selected['Total_arrival'] = pd.to_numeric(arrival_selected['Total_arrival'], errors='coerce')
arrival_selected['Total_arrival'].fillna(0, inplace=True)

expenditure_selected['Total_expend'] = pd.to_numeric(expenditure_selected['Total_expend'], errors='coerce')
expenditure_selected['Total_expend'].fillna(0, inplace=True)

expenditure_selected['Passenger_expend'] = pd.to_numeric(expenditure_selected['Passenger_expend'], errors='coerce')
expenditure_selected['Passenger_expend'].fillna(0, inplace=True)

# Merge the two datasets on 'Country' and 'Years'
merged_data = pd.merge(expenditure_selected, arrival_selected, on=['Country', 'Years'])

# Select needed columns, change type, and drop NA
merged_data = merged_data[['Country', 'Years', 'Total_expend', 'Total_arrival']]
merged_data['Years'] = pd.to_numeric(merged_data['Years'], errors='coerce')
merged_data['Total_expend'] = pd.to_numeric(merged_data['Total_expend'], errors='coerce')
merged_data['Total_arrival'] = pd.to_numeric(merged_data['Total_arrival'], errors='coerce')
merged_data.dropna(inplace=True)

# Drop rows where 'Total_arrival' is 0
merged_data = merged_data[merged_data['Total_arrival'] != 0]

merged_data['Total_arrival'] = merged_data['Total_arrival'] * 1000
merged_data['Total_expend'] = merged_data['Total_expend'] * 1000000

# Read world bank data for GDP
All_Countries_Worldbank = pd.read_csv("../data/Processed_data/All_Countries_Worldbank.csv")
All_Countries_Worldbank = All_Countries_Worldbank[['Country', 'Years', 'GDP (current US$)']]
# Replace 'UNITED STATES' with 'UNITED STATES OF AMERICA' in the 'Country' column
All_Countries_Worldbank['Country'] = All_Countries_Worldbank['Country'].replace('UNITED STATES', 'UNITED STATES OF AMERICA')

# Merge the two datasets on 'Country' and 'Years'
merged_data = pd.merge(merged_data, All_Countries_Worldbank, on=['Country', 'Years'], how='left')

# Lock axis
max_x = merged_data['Total_expend'].max()
max_y = merged_data['GDP (current US$)'].max()

# Animated plot
fig = px.scatter(
    merged_data,
    x='Total_expend',
    y='GDP (current US$)',
    size='Total_arrival',
    color='Country',
    animation_frame='Years',
    hover_name='Country',
    size_max=60,
    labels={
        'Total_expend': 'Tourism Expenditure (US$)',
        'GDP (current US$)': 'GDP (US$)',
        'Total_arrival': 'Number of Arrivals'
    },
    title='Tourism Expenditure vs GDP (2000–2022)',
)

# Fixed axes values
fig.update_layout(
    geo=dict(showframe=False),
    margin=dict(t=60, l=0, r=0, b=0),
    xaxis=dict(
        title='Tourism Expenditure (US$)',
        range=[0, max_x],
    ),
    yaxis=dict(
        title='GDP (US$)',
        range=[0, max_y]
    )
)

fig.show()

Figure 4: Bubble Plot

Other Factors Related To Tourism

The code identifies the top 10 countries with the highest tourist arrivals and uses Ridge Regression to estimate important contributing factors. A Sankey diagram then visualizes how different factors link to each country. The results show that overnight stays are major drivers for countries like Spain, Italy, and the United Kingdom. Passenger transportation is especially important for the United States and Mexico, while other tourism industries contribute notably to countries like Hungary and Poland. Although many features were included in the regression, only a few stand out clearly—most others appear less meaningful or possibly noisy.

Code

# Import libraries
import pandas as pd
import plotly.graph_objects as go
import numpy as np
from sklearn.linear_model import Ridge
from sklearn.preprocessing import StandardScaler

# Load uploaded datasets
arrival_df = pd.read_csv('../data/Processed_data/arrival.csv')
employment_df = pd.read_csv('../data/Processed_data/employment.csv')
expenditure_df = pd.read_csv('../data/Processed_data/expenditure.csv')
domestic_accommodation_df = pd.read_csv('../data/Processed_data/domestic_accommodation.csv')
# Merge datasets
merged_df = arrival_df.merge(
    employment_df, on=['Country', 'Years'], how='outer'
).merge(
    expenditure_df, on=['Country', 'Years'], how='outer'
).merge(
    domestic_accommodation_df, on=['Country', 'Years'], how='outer'
)

# Clean column names
merged_df.columns = merged_df.columns.str.replace(r'\s+\(.*?\)', '', regex=True).str.strip().str.replace(' ', '_')

# Find top arrival countries 
latest_year = arrival_df['Years'].max()
arrival_latest = arrival_df[arrival_df['Years'] == latest_year]
top_arrivals = arrival_latest[['Country', 'Total arrivals (Thousands)']].sort_values(
    by='Total arrivals (Thousands)', ascending=False
).reset_index(drop=True)
top_countries = top_arrivals['Country'].head(10)  # Top 10 countries

# Prepare feature dataset
feature_cols = [
    'Overnights_visitors', 'Same-day_visitors',
    'Accommodation_services_for_visitors', 'Food_and_beverage_serving_activities',
    'Other_accommodation_services', 'Other_tourism_industries',
    'Passenger_transportation', 'Total',
    'Travel_agencies_and_other_reservation_services_activities',
    'Passenger_transport', 'Tourism_expenditure_in_the_country',
    'Travel', 'Guests', 'Overnights', 'hotel_guests', 'hotel_overnights'
]
available_features = [col for col in feature_cols if col in merged_df.columns]

# Filter for top countries and latest year
feature_df = merged_df[(merged_df['Country'].isin(top_countries)) & (merged_df['Years'] == latest_year)]

# Prepare X and y
X = feature_df[available_features].fillna(0)
y = feature_df['Total_arrivals']

# Standardize features
scaler = StandardScaler()
X_scaled = scaler.fit_transform(X)

# Ridge Regression for feature importance
ridge_model = Ridge(alpha=1.0)
ridge_model.fit(X_scaled, y)

ridge_importance = pd.Series(ridge_model.coef_, index=available_features).sort_values(key=np.abs, ascending=False)

# Prepare chord-like (sankey) plot
top_features = ridge_importance.abs().sort_values(ascending=False).head(8)

# Labels = countries + features
labels = list(top_countries) + list(top_features.index)
n_countries = len(top_countries)
n_features = len(top_features)

# Create connection matrix
matrix = np.zeros((n_countries + n_features, n_countries + n_features))

for i, country in enumerate(top_countries):
    for j, feature in enumerate(top_features.index):
        value = feature_df.loc[feature_df['Country'] == country, feature]
        if not value.empty and not pd.isna(value.values[0]):
            matrix[i, n_countries + j] = abs(value.values[0])

# Normalize matrix for better scaling
if matrix.max() > 0:
    matrix = matrix / matrix.max()

# Build Sankey links
link = dict(source=[], target=[], value=[], color=[])
for i in range(n_countries):
    for j in range(n_features):
        if matrix[i, n_countries + j] > 0:
            link['source'].append(i)
            link['target'].append(n_countries + j)
            link['value'].append(matrix[i, n_countries + j])
            link['color'].append('rgba(150,150,250,0.5)')

# Build nodes
node = dict(
    label=labels,
    pad=15,
    thickness=20,
    line=dict(color="black", width=0.5)
)

# Plot the Sankey diagram
fig = go.Figure(data=[go.Sankey(link=link, node=node)])
fig.update_layout(title_text="Top Countries and Factors Contributing to Tourist Arrivals", font_size=10)
fig.show()

Figure 5: Sankey Plot Visualizing Factors Contributing To Tourism

To allow a closer inspection, I also created a choropleth map highlighting these top countries. In this map, you can click on a country to view key tourism-related details such as Travel Agencies, Food and Beverage Services, and Other Accommodation Services, helping to better understand the tourism infrastructure behind the visitor numbers.

Code

import pandas as pd
import plotly.graph_objects as go

# Merge data using OUTER JOIN to keep all countries
merged_df = pd.merge(employment_df, arrival_df, on=['Country', 'Years'], how='outer')

# Clean column names
merged_df.columns = merged_df.columns.str.replace(r'\s+\(.*?\)', '', regex=True).str.strip().str.replace(' ', '_')


# List of countries to add stars
highlight_countries = [
    'Mexico', 'Hungary', 'Poland', 'Italy', 'Spain', 'Türkiye', 'Denmark', 'Croatia', 'United Kingdom', 'United States'
]

# Hardcoded lat/lon for these countries
country_coords = {
    'Mexico': (23.6345, -102.5528),
    'Hungary': (47.1625, 19.5033),
    'Poland': (51.9194, 19.1451),
    'Italy': (41.8719, 12.5674),
    'Spain': (40.4637, -3.7492),
    'Türkiye': (38.9637, 35.2433),
    'Denmark': (56.2639, 9.5018),
    'Croatia': (45.1000, 15.2000),
    'United Kingdom': (55.3781, -3.4360),
    'United States': (37.0902, -95.7129)
}

# Build Dropdown Buttons
years = sorted(merged_df['Years'].dropna().unique())
data_traces = []
buttons = []

for i, year in enumerate(years):
    df_year = merged_df[merged_df['Years'] == year].copy()

    df_year['text'] = (
    "Country: " + df_year['Country'].fillna('NA') +
    "<br>Year: " + df_year['Years'].astype(str) +
     "<br>Total Arrivals: " + df_year['Total_arrivals'].fillna('NA').astype(str) + "K" +
        "<br>Overnight Visitors: " + df_year['Overnights_visitors'].fillna('NA').astype(str) + "K" +
        "<br>Same-day Visitors: " + df_year['Same-day_visitors'].fillna('NA').astype(str) + "K" +
    "<br>Food & Beverage Serving: " + df_year['Food_and_beverage_serving_activities'].fillna('NA').astype(str) + "K" +
    "<br>Other Accommodation Services: " + df_year['Other_accommodation_services'].fillna('NA').astype(str) + "K" +
    "<br>Other Tourism Industries: " + df_year['Other_tourism_industries'].fillna('NA').astype(str) + "K" +
    "<br>Passenger Transportation: " + df_year['Passenger_transportation'].fillna('NA').astype(str) + "K" +
    "<br>Travel Agencies & Reservation: " + df_year['Travel_agencies_and_other_reservation_services_activities'].fillna('NA').astype(str) + "K"
)

    # Choropleth for arrivals
    choropleth = go.Choropleth(
        locations=df_year['Country'],
        locationmode="country names",
        z=df_year['Total_arrivals'],
        text=df_year['text'],
        colorscale='YlGnBu',
        colorbar_title="Arrivals (Thousands)",
        hovertemplate="%{text}<extra></extra>",
        zmin=0,
        zmax=merged_df['Total_arrivals'].max(),
        visible=(i == 0)
    )

    # Scattergeo for stars on highlighted countries (with hardcoded lat/lon)
    lats = [country_coords[country][0] for country in highlight_countries]
    lons = [country_coords[country][1] for country in highlight_countries]

    stars = go.Scattergeo(
        lat=lats,
        lon=lons,
        mode='markers',
        marker=dict(
            size=8,
            symbol='star',
            color='black'
        ),
        hoverinfo='skip',
        visible=(i == 0)
    )

    # Append both traces for this year
    data_traces.append(choropleth)
    data_traces.append(stars)

    # Visibility settings: two traces per year
    visibility = [False] * (2 * len(years))
    visibility[2*i] = True
    visibility[2*i+1] = True

    buttons.append(dict(
        label=str(year),
        method='update',
        args=[{'visible': visibility},
              {'title': f'Tourist Arrivals and Other Relevant Indexes in {year}'}]
    ))

# Layout
layout = go.Layout(
    title=f"Tourist Arrivals and Other Relevant Indexes in {years[0]}",
    geo=dict(
        showframe=False,
        showcoastlines=True,
        projection_type='natural earth'
    ),
    height=750,
    margin=dict(l=50, r=50, t=70, b=40),
    updatemenus=[dict(
        active=0,
        buttons=buttons,
        direction="down",
        x=0.01,
        y=0.9,
        showactive=True,
        xanchor="left",
        yanchor="top"
    )]
)

# Assemble and Show
fig = go.Figure(data=data_traces, layout=layout)
fig.show()

Figure 6: Interactive Choropleth Map Displaying Global Arrival

Separately, another analysis was conducted to examine which features are most strongly correlated with total tourist arrivals. By merging datasets and calculating Pearson correlations, the code identifies the top factors influencing tourism success. The bar chart shows that overnight visitors, food and beverage services, and same-day visitors have the highest positive correlations with total arrivals, suggesting that a strong visitor experience and service offerings are closely tied to high tourism numbers. Travel and tourism expenditures also show meaningful relationships, while some factors like hotel overnights and cruise passengers have lower correlations than expected. Overall, the analysis highlights that enhancing the visitor experience beyond just lodging is crucial for attracting and sustaining tourism.

Code

import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns

# Merge on Country and Years
merged_df = (
    arrival_df
    .merge(employment_df, on=['Country', 'Years'], how='left')
    .merge(domestic_accommodation_df, on=['Country', 'Years'], how='left')
    .merge(expenditure_df, on=['Country', 'Years'], how='left')
)

# Filter to all years before 2020
df_before_2020 = merged_df[merged_df['Years'] < 2020]

# Drop rows without total arrival data
df_before_2020 = df_before_2020.dropna(subset=['Total arrivals (Thousands)'])

# Define features by their original names
features = [
    'Overnights visitors (tourists) (Thousands)',
    'Same-day visitors (excursionists) (Thousands)',
    'of which, cruise passengers (Thousands)',
    'Accommodation services for visitors (hotels and similar establishments) (Thousands)',
    'Food and beverage serving activities (Thousands)',
    'Other accommodation services (Thousands)',
    'Other tourism industries (Thousands)',
    'Passenger transportation (Thousands)',
    'Total (Thousands)',
    'Travel agencies and other reservation services activities (Thousands)',
    'Guests (Thousands)',
    'Overnights (Thousands)',
    'hotel_guests (Thousands)',
    'hotel_overnights (Thousands)',
    'Passenger transport (US$ Millions)',
    'Tourism expenditure in the country (US$ Millions)',
    'Travel (US$ Millions)'
]

# Shorter feature names
feature_name_mapping = {
    'Overnights visitors (tourists) (Thousands)': 'Overnight Visitors',
    'Same-day visitors (excursionists) (Thousands)': 'Same-day Visitors',
    'of which, cruise passengers (Thousands)': 'Cruise Passengers',
    'Accommodation services for visitors (hotels and similar establishments) (Thousands)': 'Accommodation Services',
    'Food and beverage serving activities (Thousands)': 'Food & Beverage Services',
    'Other accommodation services (Thousands)': 'Other Accommodation',
    'Other tourism industries (Thousands)': 'Other Tourism Industries',
    'Passenger transportation (Thousands)': 'Passenger Transportation',
    'Total (Thousands)': 'Total Employment',
    'Travel agencies and other reservation services activities (Thousands)': 'Travel Agencies',
    'Guests (Thousands)': 'Guests',
    'Overnights (Thousands)': 'Overnights',
    'hotel_guests (Thousands)': 'Hotel Guests',
    'hotel_overnights (Thousands)': 'Hotel Overnights',
    'Passenger transport (US$ Millions)': 'Passenger Transport Spending',
    'Tourism expenditure in the country (US$ Millions)': 'Tourism Expenditure',
    'Travel (US$ Millions)': 'Travel Spending'
}

# Compute Pearson correlation with Total arrivals
corr_matrix = df_before_2020[features + ['Total arrivals (Thousands)']]
corr_with_arrivals = corr_matrix.corr()['Total arrivals (Thousands)'].drop('Total arrivals (Thousands)')
corr_sorted = corr_with_arrivals.abs().sort_values(ascending=False)

# Prepare shorter labels for top features
top_feats = corr_sorted.head(10).index
short_labels = [feature_name_mapping[feat] for feat in top_feats]

# Plot
plt.figure(figsize=(7, 5))
sns.barplot(x=corr_with_arrivals[top_feats], y=short_labels, palette='viridis')
plt.xlabel('Correlation Coefficient\nwith Total Arrivals')
plt.title('Top 10 Features Correlated with Tourist Arrivals (Before 2020)')
plt.tight_layout()
plt.show()

Figure 7: Top Features Related To Tourism

Impact of Tourism

Tourism on GDP: Positive Effect

To explore the relationships between economic scale and tourism activity, we created an interactive parallel coordinates plot linking GDP, tourism expenditure, tourist arrivals, and years from 2010 to 2022. Each line represents a country-year observation, and users can highlight individual countries to examine their trajectories in more detail.

China exhibits massive GDP growth across the period, but its tourism expenditure and arrivals grow only modestly, showing that tourism remains a relatively small part of its overall economy.

France maintains a balanced profile, with consistently strong tourism expenditure and arrivals relative to its GDP, underscoring the important role tourism plays in its economy.

India demonstrates steady growth across GDP, expenditure, and arrivals, reflecting an emerging but still modest tourism sector.

Spain stands out as highly dependent on tourism, with tourism expenditure and arrivals forming a large share relative to its GDP, making it particularly sensitive to external shocks like the COVID-19 pandemic.

The United Kingdom shows stable GDP and moderate tourism activity, with tourism playing a meaningful but not dominant role.

Code

pd.DataFrame.iteritems = pd.DataFrame.items

# Load the data
arrival = pd.read_csv("../data/Processed_data/arrival.csv")
expenditure = pd.read_csv("../data/Processed_data/expenditure.csv")

# Select time range
arrival_2010_2022 = arrival[(arrival['Years'] >= 2010) & (arrival['Years'] <= 2022)]
expenditure_2010_2022 = expenditure[(expenditure['Years'] >= 2010) & (expenditure['Years'] <= 2022)]

# Interested countries to select
countries = ['CHINA', 'UNITED STATES OF AMERICA', 'FRANCE', 'UNITED KINGDOM', 'SPAIN', 'INDIA']

# Filter by interested countries
arrival_selected = arrival_2010_2022[arrival_2010_2022['Country'].isin(countries)]
expenditure_selected = expenditure_2010_2022[expenditure_2010_2022['Country'].isin(countries)]

countries_high = ['CHINA', 'UNITED STATES OF AMERICA']

# Filter by interested countries
arrival_selected = arrival_2010_2022[arrival_2010_2022['Country'].isin(countries_high)]
expenditure_selected = expenditure_2010_2022[expenditure_2010_2022['Country'].isin(countries_high)]

# Rename columns
arrival_selected = arrival_selected.rename(columns={'Total arrivals (Thousands)': 'Total_arrival'})
expenditure_selected = expenditure_selected.rename(columns={
    'Tourism expenditure in the country (US$ Millions)': 'Total_expend',
    'Passenger transport (US$ Millions)': 'Passenger_expend',
    'Travel (US$ Millions)': 'Travel_expend'
})

# Ensure numeric and handle missing values (for both datasets)
arrival_selected['Total_arrival'] = pd.to_numeric(arrival_selected['Total_arrival'], errors='coerce')
arrival_selected['Total_arrival'].fillna(0, inplace=True)

expenditure_selected['Total_expend'] = pd.to_numeric(expenditure_selected['Total_expend'], errors='coerce')
expenditure_selected['Total_expend'].fillna(0, inplace=True)

expenditure_selected['Passenger_expend'] = pd.to_numeric(expenditure_selected['Passenger_expend'], errors='coerce')
expenditure_selected['Passenger_expend'].fillna(0, inplace=True)

# Merge the two datasets on 'Country' and 'Years'
merged_data = pd.merge(expenditure_selected, arrival_selected, on=['Country', 'Years'])

# Select needed columns, change type, and drop NA
merged_data = merged_data[['Country', 'Years', 'Total_expend', 'Total_arrival']]
merged_data['Years'] = pd.to_numeric(merged_data['Years'], errors='coerce')
merged_data['Total_expend'] = pd.to_numeric(merged_data['Total_expend'], errors='coerce')
merged_data['Total_arrival'] = pd.to_numeric(merged_data['Total_arrival'], errors='coerce')
merged_data.dropna(inplace=True)

# Drop rows where 'Total_arrival' is 0
merged_data = merged_data[merged_data['Total_arrival'] != 0]

merged_data['Total_arrival'] = merged_data['Total_arrival'] * 1000
merged_data['Total_expend'] = merged_data['Total_expend'] * 1000000

# Read world bank data for GDP
All_Countries_Worldbank = pd.read_csv("../data/Processed_data/All_Countries_Worldbank.csv")
All_Countries_Worldbank = All_Countries_Worldbank[['Country', 'Years', 'GDP (current US$)']]
# Replace 'UNITED STATES' with 'UNITED STATES OF AMERICA' in the 'Country' column
All_Countries_Worldbank['Country'] = All_Countries_Worldbank['Country'].replace('UNITED STATES', 'UNITED STATES OF AMERICA')

# Merge the two datasets on 'Country' and 'Years'
merged_data = pd.merge(merged_data, All_Countries_Worldbank, on=['Country', 'Years'], how='left')

# Create a numerical encoding for the 'Country' column
merged_data['Country'] = merged_data['Country'].astype('category')
merged_data['Country_Code'] = merged_data['Country'].cat.codes

# Plotly
fig = px.parallel_coordinates(merged_data, color="Country_Code",
                              dimensions=['GDP (current US$)', 'Total_expend', 'Total_arrival', 'Years'],
                              color_continuous_scale=px.colors.qualitative.Set1,
                              color_continuous_midpoint=1, range_color = [0, 1.5],
                              title = "Relations For Inbound Arrivals, Expenditure, and GDP (Higher GDP Tier)", width = 800, height = 600)
fig.update_layout(
    margin=dict(t=100, l=50, r=50, b=50),
    title=dict(y=0.95),
    coloraxis_colorbar=dict(
        tickvals=list(range(len(merged_data['Country'].cat.categories))),  # Match ticks with country codes
        ticktext=merged_data['Country'].cat.categories.tolist(),  # Show country names instead of numeric codes
        title="Countries"
    ),
    showlegend=True  # Ensure the legend is visible
)
fig.show()

Figure 8: Parallel Coordinate Plot: Higher Tier

Code

countries_low = ['FRANCE', 'UNITED KINGDOM', 'SPAIN', 'INDIA']

# Filter by interested countries
arrival_selected = arrival_2010_2022[arrival_2010_2022['Country'].isin(countries_low)]
expenditure_selected = expenditure_2010_2022[expenditure_2010_2022['Country'].isin(countries_low)]

# Rename columns
arrival_selected = arrival_selected.rename(columns={'Total arrivals (Thousands)': 'Total_arrival'})
expenditure_selected = expenditure_selected.rename(columns={
    'Tourism expenditure in the country (US$ Millions)': 'Total_expend',
    'Passenger transport (US$ Millions)': 'Passenger_expend',
    'Travel (US$ Millions)': 'Travel_expend'
})

# Ensure numeric and handle missing values (for both datasets)
arrival_selected['Total_arrival'] = pd.to_numeric(arrival_selected['Total_arrival'], errors='coerce')
arrival_selected['Total_arrival'].fillna(0, inplace=True)

expenditure_selected['Total_expend'] = pd.to_numeric(expenditure_selected['Total_expend'], errors='coerce')
expenditure_selected['Total_expend'].fillna(0, inplace=True)

expenditure_selected['Passenger_expend'] = pd.to_numeric(expenditure_selected['Passenger_expend'], errors='coerce')
expenditure_selected['Passenger_expend'].fillna(0, inplace=True)

# Merge the two datasets on 'Country' and 'Years'
merged_data = pd.merge(expenditure_selected, arrival_selected, on=['Country', 'Years'])

# Select needed columns, change type, and drop NA
merged_data = merged_data[['Country', 'Years', 'Total_expend', 'Total_arrival']]
merged_data['Years'] = pd.to_numeric(merged_data['Years'], errors='coerce')
merged_data['Total_expend'] = pd.to_numeric(merged_data['Total_expend'], errors='coerce')
merged_data['Total_arrival'] = pd.to_numeric(merged_data['Total_arrival'], errors='coerce')
merged_data.dropna(inplace=True)

# Drop rows where 'Total_arrival' is 0
merged_data = merged_data[merged_data['Total_arrival'] != 0]

merged_data['Total_arrival'] = merged_data['Total_arrival'] * 1000
merged_data['Total_expend'] = merged_data['Total_expend'] * 1000000

# Read world bank data for GDP
All_Countries_Worldbank = pd.read_csv("../data/Processed_data/All_Countries_Worldbank.csv")
All_Countries_Worldbank = All_Countries_Worldbank[['Country', 'Years', 'GDP (current US$)']]
# Replace 'UNITED STATES' with 'UNITED STATES OF AMERICA' in the 'Country' column
All_Countries_Worldbank['Country'] = All_Countries_Worldbank['Country'].replace('UNITED STATES', 'UNITED STATES OF AMERICA')

# Merge the two datasets on 'Country' and 'Years'
merged_data = pd.merge(merged_data, All_Countries_Worldbank, on=['Country', 'Years'], how='left')

# Create a numerical encoding for the 'Country' column
merged_data['Country'] = merged_data['Country'].astype('category')
merged_data['Country_Code'] = merged_data['Country'].cat.codes

# Plotly
fig = px.parallel_coordinates(merged_data, color="Country_Code",
                              dimensions=['GDP (current US$)', 'Total_expend', 'Total_arrival', 'Years'],
                              color_continuous_scale=px.colors.qualitative.Set1,
                              color_continuous_midpoint=1, range_color = [0, 3.5],
                              title = "Relations For Inbound Arrivals, Expenditure, and GDP (Lower GDP Tier)", width = 800, height = 600)
fig.update_layout(
    margin=dict(t=100, l=50, r=50, b=50),
    title=dict(y=0.95),
    coloraxis_colorbar=dict(
        tickvals=list(range(len(merged_data['Country'].cat.categories))),  # Match ticks with country codes
        ticktext=merged_data['Country'].cat.categories.tolist(),  # Show country names instead of numeric codes
        title="Countries"
    ),
    showlegend=True  # Ensure the legend is visible
)
fig.show()

Figure 9: Parallel Coordinate Plot: Lower Tier

Tourism and Employment

Tourism growth has consistently driven job creation across all six markets. China and India’s visitor numbers surged 4–6× since 2000, fueling substantial employment gains even in capital-intensive sectors. France and the U.K., despite more modest arrival increases, generated proportionally greater jobs, showcasing labor-rich tourism models. In the U.S. and Spain, tourism and employment rose hand-in-hand through 2019 and rebounded strongly after the 2020 downturn. Overall, rising arrivals translate directly into more tourism-sector jobs, confirming a clear, positive impact of tourism on employment.

Code

import pandas as pd
import matplotlib.pyplot as plt

arrival = pd.read_csv('../data/Processed_data/arrival.csv')
employment = pd.read_csv('../data/Processed_data/employment.csv')

target = [
    'TAIWAN PROVINCE OF CHINA',
    'UNITED KINGDOM',
    'UNITED STATES OF AMERICA',
    'INDIA',
    'FRANCE',
    'SPAIN'
]
rename_map = {
    'TAIWAN PROVINCE OF CHINA': 'China',
    'UNITED KINGDOM': 'United Kingdom',
    'UNITED STATES OF AMERICA': 'United States',
    'INDIA': 'India',
    'FRANCE': 'France',
    'SPAIN': 'Spain'
}

arr = arrival[arrival['Country'].isin(target)].copy()
emp = employment[employment['Country'].isin(target)].copy()
arr['Country'] = arr['Country'].replace(rename_map)
emp['Country'] = emp['Country'].replace(rename_map)

arr_pivot = arr.pivot(index='Years', columns='Country', values='Total arrivals (Thousands)')
emp_pivot = emp.pivot(index='Years', columns='Country', values='Total (Thousands)')

countries = ['China', 'United Kingdom', 'United States', 'India', 'France', 'Spain']
arr_index = pd.DataFrame(index=arr_pivot.index)
emp_index = pd.DataFrame(index=emp_pivot.index)
base_years = {}

for country in countries:
    arr_years = set(arr_pivot.index[arr_pivot[country].notna()])
    emp_years = set(emp_pivot.index[emp_pivot[country].notna()])
    common_years = sorted(arr_years & emp_years)
    if not common_years:
        continue
    base = common_years[0]
    base_years[country] = base
    arr_index[country] = arr_pivot[country] / arr_pivot.loc[base, country] * 100
    emp_index[country] = emp_pivot[country] / emp_pivot.loc[base, country] * 100

ymin = min(arr_index.min().min(), emp_index.min().min()) * 0.9
ymax = max(arr_index.max().max(), emp_index.max().max()) * 1.1

fig, axes = plt.subplots(2, 3, figsize=(10, 6), sharex=True, sharey=True)
axes = axes.flatten()

for ax, country in zip(axes, countries):
    base = base_years.get(country, None)
    ax.plot(arr_index.index, arr_index[country], label='Arrivals', color='tab:blue', linewidth=1.5)
    ax.plot(emp_index.index, emp_index[country], label='Employment', color='tab:red', linestyle='--', linewidth=1.5)
    ax.set_title(country)
    ax.set_xlabel('Year')
    ax.set_ylabel('Index (100 = base)')
    ax.set_ylim(ymin, ymax)
    ax.grid(True, linestyle=':', linewidth=0.5)
    ax.legend(loc='upper left', fontsize='small')

plt.suptitle('Indexed Growth: Tourism vs. Employment', fontsize=16)
plt.tight_layout(rect=[0, 0.03, 1, 0.95])
plt.show()

Figure 10: Indexed Growth: Tourism vs. Employment

Conclusion

Our analysis reveals that tourism success is driven more by cultural appeal, service quality, and accessibility than by pure economic size. Strong tourism sectors rely on creating rich visitor experiences, not just wealth. In return, tourism acts as a major engine for economic growth—boosting employment and strengthening economies. Countries that nurture both domestic and international tourism stand out not just as travel hubs, but also as more adaptable and prosperous economies.